⚙
AI-generated deep summary
by claude@2026-07, 2026-07-03
· read from full text
ⓘ
The paper studies morphological effects of 53 small-molecule compounds targeting six programmed cell death mechanisms across multiple concentrations in MCF7 cells, using the Cell Painting assay combined with single-cell morphological profiling and feature extraction. It compares single-cell versus aggregated analysis strategies, reporting that self-supervised DINO embeddings on single-cell data capture high-resolution morphological patterns, and that an energy-distance metric helps quantify perturbation strength and filter relevant profiles. Focused analysis of apoptosis-inducing compounds shows concentration-dependent biological heterogeneity in single-cell data that is not apparent in aggregated profiles, while multi-class models for six programmed cell death mechanisms perform best with aggregated features for supervised classification (with up to ~89.97% F1) despite single-cell advantages for unsupervised exploration. A key caveat is that the experiments are limited to MCF7 cells and morphological readouts, rather than direct mechanistic validation, and the authors note morphological limitations related to how “cell death” is treated. The paper does not explicitly discuss endometriosis or adenomyosis; it was included in the corpus via a keyword match in the upstream search index.
Abstract
Analysis at the single-cell level is a powerful approach to study biological processes and responses to perturbations. However, its application in morphological profiling with phenomics remains underexplored. Here, we use the Cell Painting assay to investigate morphological effects of 53 small molecule compounds, associated with six distinct programmed cell death mechanisms, across six concentrations in MCF7 cells. To compare single-cell and aggregated analysis strategies, we conduct both supervised and unsupervised evaluations aimed at identifying features linked to programmed cell death. We apply an energy distance as a metric to quantify morphological perturbation strength, enabling efficient filtering. Among three tested feature extraction methods, self-supervised DINO embeddings applied to single-cell data captured high-resolution morphological patterns. Focused analyses of apoptosis-inducing compounds revealed biological heterogeneity attributable to specific molecular targets and concentration-dependent effects, which were not apparent in aggregated profiles. In contrast, multi-class classification models for the six programmed cell death mechanisms trained on single-cell features achieved F1 scores of 79.86%, while models trained on aggregated features reached F1 scores of up to 89.97%. Our results highlight the advantages of single-cell data for unsupervised exploration and show that aggregated representations yield more robust and accurate performance in supervised models.
Full text
3,079 characters
· extracted from
oa-doi-fallback
· click to expand
Abstract
Analysis at the single-cell level is a powerful approach to study biological processes and responses to perturbations. However, its application in morphological profiling with phenomics remains underexplored. Here, we use the Cell Painting assay to investigate morphological effects of 53 small molecule compounds, associated with six distinct programmed cell death mechanisms, across six concentrations in MCF7 cells. To compare single-cell and aggregated analysis strategies, we conduct both supervised and unsupervised evaluations aimed at identifying features linked to programmed cell death. We apply an energy distance as a metric to quantify morphological perturbation strength, enabling efficient filtering. Among three tested feature extraction methods, self-supervised DINO embeddings applied to single-cell data captured high-resolution morphological patterns. Focused analyses of apoptosis-inducing compounds revealed biological heterogeneity attributable to specific molecular targets and concentration-dependent effects, which were not apparent in aggregated profiles. In contrast, multi-class classification models for the six programmed cell death mechanisms trained on single-cell features achieved F1 scores of 79.86%, while models trained on aggregated features reached F1 scores of up to 89.97%. Our results highlight the advantages of single-cell data for unsupervised exploration and show that aggregated representations yield more robust and accurate performance in supervised models.
Competing Interest Statement
J.C.P. and O.S. declare ownership in Phenaros Pharmaceuticals.
Footnotes
Replaced cell death with programmed cell death throughout to reflect morphological limitations. Expanded Methods: Detailed imaging setup, plate layout, field-of-view, and channel structure. Described QC pipeline, segmentation procedures, illumination correction, and sampling strategy. Added full DINO training details and hyperparameters. New Analyses: Dose-response analysis based on nuclear counts. k*-distribution and skewness metrics for morphological heterogeneity. Pairwise e-distances between compound concentrations. Included mAP (mean average precision) alongside grit and e-distance for filtering and evaluation. Data Processing Updates: Matched cell sampling across feature extractors (DINO, CellProfiler, DeepProfiler). Re-ran DeepProfiler and CellProfiler with stratified downsampling for comparability. Figures and Supplement: Reordered and renumbered all supplementary figures. Improved figure legibility (scale bars, color palette, legends). Corrected confusion matrix color scales and labels. Added compound counts and MOA references to legends and tables. Public Resources: Updated GitHub with complete preprocessing scripts, QC pipeline, and evaluation notebooks. Added missing links and fixed execution errors in notebooks. Updated Figshare with consistent test splits, QC flags, and feature embeddings. Drug Metadata: Added literature-backed references for all compound MoAs. Clarified pleiotropy and class imbalance treatment in dataset.
Text is read by the "Ask this paper" AI Q&A widget below.
Extraction quality varies by source — PMC NXML preserves structure
cleanly, OA-HTML may include some navigation residue, and OA-PDF can
have broken hyphenation. The publisher copy
(via DOI)
is the canonical version.